Can Interest Rate Dynamics Save Structural Models?1

1I would like to aknowledge the Ente Luigi Einaudi for generously funding this research during my stay as a research fellow in Rome during the academic year 2003-2004. The results for structural models for corporate bonds in the recent empirical literature are at best mixed; in general, such models generate credit spreads considerably below empirically observed levels. However, several authors have found that, for realistic parameter values, such models do seem able to generate levels for default probabilities in line with empirically observed default frequencies. This seems to suggest that the link between the physical and risk-neutral dynamics in the existing structural models might be too tight, such that a model can be calibrated to either default frequencies or credit spreads, but not to both. In this paper the structural model of Collin-Dufresne and Goldstein (2001) is extended to incorporate a two-factor essentially affine term structure model. The results, generated along the lines of Huang and Huang (2003), strongly suggest that allowing for essentially affine dynamics significantly improves the ability of a structural model to generate credit spreads of realistically high levels.